Much has been written on the use of observational studies in epidemiological research. It is necessary to apply many of the same epidemiologic techniques used in conventional observational study designs to the clinical data bank. Work on determining which epidemiologic approaches are most appropriate for use with clinical data banks is continuing. In the past, this project has focused upon quality assurance methods; at present, the emphasis is on epidemiological considerations in the analysis and interpretation of data bank results. Aspects under study include 1) generalizability of results from multicentered hospital-based cases, 2) methods for detecting inter-center variation and determining conditions when pooling of results is appropriate, and 3) use of the data bank as a source of cases for case-control studies.